Free guide · Oct 5, 2026
Strata: run a 125B AI model on your gaming PC
A step-by-step setup guide with requirements, checks and common problems.
The short version
Strata is a free, open source tool that runs Qwen3.8-Flash-Next, a 125 billion parameter model that normally needs a server, on one NVIDIA graphics card and your PC's RAM. The README measures 53 to 93 tokens per second on an RTX 5070 with 12 GB of VRAM. That's faster than you can read.
•What you need: an NVIDIA RTX 20, 30, 40 or 50 card with 12 GB of VRAM or more, 64 GB of RAM (48 GB works for the two smallest sizes, 32 GB for the coding version), about 80 GB of free disk space, and Windows 10/11 or Linux.
•What you get: a local chat app in your browser, plus an OpenAI and Anthropic compatible API on your own PC that your apps and coding agents can use.
What's in the full guide
- 01How fast is it?
- 02Will it fit on my PC?
- 03Step 1: Download it
- 04Step 2: Double-click START-HERE.bat
- 05Step 3: The first start
- 06Step 4: Use it
- 07Step 5: Get a free speed boost
- 08If you have less RAM or VRAM
- 09Where everything is stored
- 10Honest notes
- 11Where to go next